Steps to reproduce:

1. Add a VMware ESXi host (or vcenter server) to zabbix, and enable VMware Guest to automatically create the VMs as registered hosts. 1x VM exists for testing called `mytestvm.domain.local`
2. Verify the host in zabbis is polling, had to wait 1 hour before statistics were collected.
3. RDP to the VM in question, launch task manager and observe the system boot-time counter.  147 days+
4. Initiate a separate "ping -t" command to the host by IP address. Confirmed stable reply of less than 1ms.
5. Perform a VMware vMotion of the Host VM from ESXi host 1 to ESXi host 2. Note RDP session remains connected, and the system uptime remains counting (still several weeks).
6. Check Zabbix, a new PROBLEM record is created by the 'VMware Guest / VM has been restarted` trigger.  THIS IS A FALSE POSITIVE
   Expression reads:
   (between(last(/mytestvm.domain.local/vmware.vm.guest.osuptime[\{$VMWARE.URL},\{$VMWARE.VM.UUID}]),1,10m)=1 or between(last(/mytestvm.domain.local/vmware.vm.uptime[\{$VMWARE.URL},\{$VMWARE.VM.UUID}]),1,10m)=1) and last(/mytestvm.domain.local/vmware.vm.powerstate[\{$VMWARE.URL},\{$VMWARE.VM.UUID}]) = 1

Root-Cause:

• During a vMotion operation, the VM's VM-world-ID is recreated on the destination host, which is the direct of this Zabbix template, misinterpreting the vmotion-event as a vm-restart. 

• The item `vmware.vm.uptime` metric (which tracks the VM's uptime from the single-hypervisor's perspective) is designed to live only as long as that VM is alive on that host. Thus `this metric is supposed to reset-to-zero` (evidenced by my screenshot below) during the vMotion process, even though the guest OS uptime (as also evidenced) remains counting.

• Zabbix's trigger logic is designed to detect `VM restarts` based on the wrong metric. vmware.vm.uptime during vMotion triggers a false positive.

• The expression in the template trigger reads:
  `
  (between(last(/VMware Guest/vmware.vm.guest.osuptime[\{$VMWARE.URL},\{$VMWARE.VM.UUID}]),1,10m)=1 or between(last(/VMware Guest/vmware.vm.uptime[\{$VMWARE.URL},\{$VMWARE.VM.UUID}]),1,10m)=1) and last(/VMware Guest/vmware.vm.powerstate[\{$VMWARE.URL},\{$VMWARE.VM.UUID}]) = 1
  `
  This should be revised to just include:
  `
  between(last(/VMware Guest/vmware.vm.guest.osuptime[\{$VMWARE.URL},\{$VMWARE.VM.UUID}]),1,10m)=1
  and last(/VMware Guest/vmware.vm.powerstate[\{$VMWARE.URL},\{$VMWARE.VM.UUID}]) = 1
  ` 

Evidence: